Question

我正在尝试使用我的图像数据集训练具有张量流的分类器。而且我不断收到此错误。

ValueError: Cannot feed value of shape (18,) for Tensor 'targets/Y:0', which has shape '(?, 2)'

我真的是机器学习和python的新手，我试图在所有地方搜索这种类型的错误，但找不到自己的解决方案。请帮忙。我正在使用巨大的图像数据集和标签，并尝试使用它训练分类器。

我的代码在下面：

import numpy as np 
import os
from keras.models import Sequential
import scipy.io as cio
import matplotlib.image as mpg
import matplotlib.pyplot as plt
from random import shuffle
import tflearn
from tflearn.layers.conv import conv_2d, max_pool_2d
from tflearn.layers.core import input_data, dropout, fully_connected
from tflearn.layers.estimator import regression
import tensorflow as tf
import cv2
from tqdm import tqdm

image = "/data/image/1/1101"
IMG_SIZE = 50
training_data = []
testing_data = []
MODEL_NAME = 'Classification'
LR = 1e-3




def rgb2gray(rgb):
    return np.dot(rgb[...,:3], [0.299, 0.587, 0.114])

def labelling(mylabel):
  if mylabel == '-1':
    return -1 #'uncertain'
  elif mylabel == '1':
    return 1 #'Front'
  elif mylabel == '2':
    return 2 #'Rear'
  elif mylabel == '3':
    return 3 #'Side'
  elif mylabel == '4':
    return 4 #'Front-Side'
  elif mylabel == '5':
    return 5 #'Read-Side'

for root, _, files in os.walk(image):
  cdp = os.path.abspath(root)

  for f in files:
    name,ext = os.path.splitext(f)
    if ext == ".jpg":
      cip = os.path.join(cdp, f)
      ci = mpg.imread(cip)
      ci = rgb2gray(ci)
      # images = cv2.resize(ci,(IMG_SIZE,IMG_SIZE))

      # images = rgb2gray(images)
      images = cv2.cv2.resize(ci,(IMG_SIZE,IMG_SIZE))
      images = np.array(images)
      # print(images.shape)
      #images = images / 255
      # print(images)
      # count = count + 1 
      label = cip.replace('image','label')
      label = label.replace('.jpg', '.txt')


      lines =[line.rstrip('\n') for line in open(label)]
      # print(lines[0])
      lines = np.array(lines)
      # print(lines)


      training_data.append((images,labelling(lines[0])))
      shuffle(training_data)
      np.save('training_data.npy',training_data)
      # print(training_data)


for img in tqdm(os.listdir(image)):
  path = os.path.join(image,img)
  img_num = img.split('.')[0]
  testing_data.append((np.array(images),img_num))
  shuffle(testing_data)


  train = training_data[:-5]

  test = testing_data[-5:]

  X_train = np.array([i[0] for i in training_data]).reshape(-1, IMG_SIZE, IMG_SIZE, 1)

  y_train = [i[1] for i in training_data]

  X_test = np.array([i[0] for i in testing_data]).reshape(-1, IMG_SIZE, IMG_SIZE, 1)
  print("XTEST",X_test)
  y_test = [i[1] for i in testing_data]
  print(y_test)



tf.reset_default_graph()
convnet = input_data(shape=[None,IMG_SIZE,IMG_SIZE,1],name='input')
#shape=[None, IMG_SIZE, IMG_SIZE, 1],
convnet = conv_2d(convnet, 32, 5, activation='relu')
convnet = max_pool_2d(convnet, 5)
convnet = conv_2d(convnet, 64, 5, activation='relu')
convnet = max_pool_2d(convnet, 5)
convnet = conv_2d(convnet, 128, 5, activation='relu')
convnet = max_pool_2d(convnet, 5)
convnet = conv_2d(convnet, 64, 5, activation='relu')
convnet = max_pool_2d(convnet, 5)
convnet = conv_2d(convnet, 32, 5, activation='relu')
convnet = max_pool_2d(convnet, 5)
convnet = fully_connected(convnet, 1024, activation='relu')
convnet = dropout(convnet, 0.8)
convnet = fully_connected(convnet, 2, activation='softmax')
convnet = regression(convnet, optimizer='adam', learning_rate=LR, loss='categorical_crossentropy', name='targets')
model = tflearn.DNN(convnet, tensorboard_dir='log', tensorboard_verbose=0)
model.fit({'input': X_train}, {'targets': y_train}, n_epoch=10,
          validation_set=({'input': X_test}, {'targets': y_test}),
          snapshot_step=500, show_metric=True, run_id=MODEL_NAME)
#({'input': X_test}, {'targets': y_test})

Tensorflow，TF ValueError：ValueError：无法为形状为（（？，2）'的Tensor'targets / Y：0'输入形状（18，）的值

0 个答案: